Amazon's AgentCore Browser Tool Lets AI Agents Automate Legacy Web Apps

Amazon Bedrock AgentCore Browser Tool tackles a persistent enterprise problem: legacy web applications that rely on server-side HTML rendering and lack modern APIs. Traditional RPA struggles with multi-step workflows, dynamic form validation, and session-dependent state. The tool provides a fully managed, cloud-based Chromium browser that AI agents can control via Playwright over WebSocket-based Chrome DevTools Protocol (CDP). Target applications only need HTTP/HTTPS access, regardless of original browser compatibility. Combined with Strands Agents for model-driven orchestration, it scales from single-step automations to complex multi-agent workflows, integrating with Amazon Bedrock foundation models through the AgentCore runtime.

Security and compliance are central: each session is isolated, protected by AWS IAM policies, and fully auditable, preserving human oversight for regulated industries like insurance, healthcare, and finance. The accompanying post details a reference implementation of an AI-powered digital worker, covering architecture, design decisions, and an enterprise blueprint with Terraform deployment. Complete source code is available on GitHub, enabling organizations to deploy and customize the automation for their own legacy systems. This approach lets enterprises reduce costly manual data entry and errors without ripping out decades-old infrastructure.